News for hackers. Stuff that matters.
Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by DrNuke on Tuesday December 01, 2020 @06:10AM (#25264235)
It was mainly an engineering issue: materials, manufacturing and assembling. It is too big to fail right now, and most issues have been solved. It never was as simple as webinars from scientists would make it. The funniest joke within nuclear eng insiders is the marvels of paper reactors pretty always make real reactors just seem glorified boilers.
Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by cromwellian on Tuesday December 01, 2020 @06:28AM (#25264330)
ITER's too big the fail. But again, the parameters were known. The magnets give you the field strength, the field strength tells you the radius, the radius^3 gives you the cost, and also, some measure of the construction time.

This is one of the great things about Elon Musk -- he looks at things from first principles. When it looked like he needed 1000 Gigafactories to reach his goal, he went back to the drawing board on decades of battery manufacturing, to come up with the Terafactory.

When carbon composite rockets looked to be too hard to manufacture, and have too many other detriments, despite the clear wins in strength-to-weight, he dropped it and went with stainless steel. I'm pretty sure if he was staring at the basic Tokamak equations, he would have long again saw the economic and time dependence on R, and the massive benefits to improving B, and that would have driven him to look for alternatives.

Too many researchers engage in sunk cost / dollar auctions, doubling down on obvious wrong approaches, because they've already spent so much.

This is also why DeepMind won CASP again, because rather than try to iteratively improve their 2018 result, they dropped it as a dead end and restarted with a completely new approach, even though the 2018 approach was also a revolutionary breakthrough.

What I'm getting at is, being able to let go of your previous projects if they are stalling, with huge cost over runs, and a mediocre future return, is a value we should seek in fusion research, and elsewhere.

Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by DrNuke on Tuesday December 01, 2020 @06:42AM (#25264382)
Ok but ITER was much more than engineering: soft diplomacy, supranational bureaucracy, manufacturing modernisation. Musk comes from a different era and a different environment but, again, launching rockets successfully, with humans in them safely going and coming back, was achieved well before him. Stable nuclear fusion is still unclaimed as a milestone, even if sustainable nuclear fusion might well come from different, smaller approaches the likes we are seeing from United Kingdom and South Korea.
Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by pfdietz on Tuesday December 01, 2020 @05:09PM (#25270668)
Going back to first principles is how you conclude DT fusion is a dead end. And this has been known for decades.

http://orcutt.net/weblog/wp-content/uploads/2015/08/The-Trou...

(That prescient article has stood the test of time, except that advanced fuels were mostly ruled out also.)

Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by cromwellian on Wednesday December 02, 2020 @03:43AM (#25274479)
Except the article is wrong, because it makes fundamental assumptions about the vacuum wall and lithium blanket, and even the superconductors. The ARC reactor, for example, doesn't use a solid sandwich style blanket, it uses a volumetric blanket of FLiBe, a liquid that acts as the moderator that absorbs neutrons, the breeder, the magnetic shield, and the heat exchanger -- and has proven to work in molten salt reactor prototypes. And unlike the designs he claims that are "essentially unchanged", this significantly reduces the cost and mass of the reactor, reduces the so-called "danger" of lithium, and leverages the fact that high energy fusion neutrons travel further, as it increases the efficacy of the liquid blanket.

The claim that neutrons will damage the metal structure is also handled by many aspects of the design: easy reparability thanks to modular, segmented magnets, being completely submerged in the liquid blanket, low activation steel for structure, and the pulse design which runs the reactor in a way that reduces overall temperature of the walls.

Plus, he brings up the ole aneutronic fusion dream, except that creating a confined Boron-11 plasma is much more difficult, outside the realm of tokamaks, and typically proposes using laser internal confinement fusion at the Petawatt scale.

More than 20 papers were published on ARC, heavily vetted in journals like Nature, and pretty much all of the peer review concluded the design will work, and if it does work, then it will be much cheaper than ITER, simpler than ITER, and thus refute your paper.

We'll know in about 4 years.

Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by pfdietz on Wednesday December 02, 2020 @08:12AM (#25276007)
No. The article makes no assumptions about superconductors. The argument makes no assumption even about plasma confinement. It's all about the limits on areal power density at the reactor walls.

The conclusion he reaches, that fusion reactors will have a volumetric power density at least an order of magnitude lower than a fission reactor, is not contradicted by ARC. The volumetric power density of ARC (from the ARC paper I linked to elsewhere in these comments) is 0.5 MW(th)/m^3 (counting the volume of the entire reactor, not just the plasma) vs. 20 MW(th)/m^3 for the primary reactor vessel of a PWR -- a factor of 40 worse.

His suggestion to look at aneutronic plasmas didn't pan out, as you note, but that doesn't save DT fusion. I'll note that it's his PhD student Todd Rider who shot down the non-Maxwellian schemes that aneutronic fuels would have needed. After that, Lidsky switched to working on fission for the rest of career and life.

Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by cromwellian on Wednesday December 02, 2020 @04:43PM (#25282049)
The latest ARC claims from Whyte's last presentation was 3MW/m^3 density, and that's with currently achievable magnets, not those people believe are ultimately achievable. Counting just the PWR reactor vessel, and not the entire containment facility, seems a huge cheat.

We're talking economics of energy. PWR Fission reactors have enormous footprints, averaging 1 square mile. An ARC reactor wouldn't need anywhere need that size. The very nature of using pressurized water means your containment facility has to be a factor of 1000 larger than the reactor vessel to handle a flash boil over.

Current costs to build a 1GW fission plant are $6-9 billion. Predicted costs to build a 200MW ARC reactor are $4-5 billion, so the ARC reactor costs ~5x more at the high end, and 77% more at the low end, for the same output (e.g. build 5 ARC reactors to equal one fission plant). The economics are still viable, and your costs for storage, disposal, and de-comissioning are far lower, as is your land footprint. Note that these ARC costs are for the first reactor, include the sunk development costs, and not the economies of scale that would occur from dropping HTS costs.

> It's all about the limits on areal power density at the reactor walls.

No, he makes lots of assumptions about how the magnets are shielded, how breeding works, and how heat transfer works. All of these are addressed by FLiBe, there's no "structure" to the heat blanket to damage, it acts simultaneously as radioactive shielding for the magnets, heat transfer, and a breeder. The "blanket" in ARC isn't a surface sandwiched with a super structure, it's a volume that's continuously pumped and recirculated.

He also makes the implicit assumption that reactor wall damage is a huge problem. Instead of seeing it as a problem, you should see it as a solution, like an ablative heat shield on a rocket or capsule. As long as the cost of maintenance or replacement is low, and reactor downtime is limited, it is not a problem. Current estimates for ARC, which is designed to be modular, so it can be easily maintained, is that the reactor walls will last about 1 year.

He doesn't even mention the biggest unsolved engineering problem, removing the "ash" (Helium) that accumulates in the plasma.

His article has all the trappings of reusable rocket naysayers in the 80s and 90s, who claimed reusables could never work, because carrying wings, landing legs, heat shielding, and structural reinforcement to make rapid reusability would cut down on your usable payload, and thus doom reusable rockets to always be substantially inferior. And this is because of an obsession with a single variable, like power density, or payload fraction, and ignoring the end to end economics of the entire system, including safety margins, maintenance, footprint, etc.

And that's why I believe ARC will prove Lidsky wrong.

Re:South Korea's fusion device KSTAR runs for 20 seconds at ... by pfdietz on Wednesday December 02, 2020 @05:30PM (#25282606)
> The latest ARC claims from Whyte's last presentation was 3MW/m^3 density,

That's per volume of PLASMA, not volume of REACTOR. Since we're talking about the size of the things we have to pay for (the reactor) the latter is appropriate.

> Counting just the PWR reactor vessel, and not the entire containment facility, seems a huge cheat.

Fusion reactors will also require containment vessels, to prevent escape of tritium in accidents (unless you're ok with an accident rendering ground water in a large area around the reactor undrinkable for decades.) The containment vessel will have to be large enough to contain all the cryogens heated to accident temperature. Unlike in a PWR like the AP1000, these volatiles cannot be condensed by water cooling. In ITER, for example, complete volatilization of the cryogens would requires several times the volume of the AP1000 containment building.

The ARC design adds to accident fun by having a large volume of titanium hydride shielding near the molten salt. TiH2 decomposes to Ti and hydrogen gas at 450 C, well below the temperature of the molten salt blanket. Fully decomposed, this TiH2 would release 15 tonnes of hydrogen, occupying 180,000 cubic meters at STP.

Beyond that, the containment building around a fusion reactor will have to mitigate tritium leakage by many orders of magnitude. This will be expensive (in particular, polymer seals cannot be used on penetrations as they permeate tritium.)

ARC will also require a large volume for disassembly of the reactor for vacuum vessel replacement, and for remote control demolition of activated vacuum vessels for disposal. All that volume will become contaminated and off limits to human access.

> PWR Fission reactors have enormous footprints, averaging 1 square mile. An ARC reactor wouldn't need anywhere need that size.

Really good farmland in the US is less than $10K/acre, or about $6M/square mile. Cheap range land is an order of magnitude less; really useless land another order of magnitude cheaper. The cost of land there is a very minor part (0.1%) of the cost of the power plant.

https://www.howmuchisit.org/acre-of-land-cost/

> No, he makes lots of assumptions about how the magnets are shielded, how breeding works, and how heat transfer works.

Again, you have not understood his argument. His argument would apply even if you had 100T magnets made of unobtainium and a plasma configuration with beta = 1. All that he requires is an upper limit on how much energy per area can be transmitted through the surface of the reactor. This is a function of the materials of the first wall, and has absolutely nothing to do with magnets or breeding.

> Predicted costs to build a 200MW ARC reactor are $4-5 billion

If you are referring to the cost estimate from the arxiv paper, that is just the estimated cost of the reactor itself, not the power plant containing the reactor. I will admit I found the estimation methodology in that paper very cursory, so it's possible they were overestimating the cost (80% of the cost was the stainless steel magnet support structure.)

A fission reactor core is much cheaper than that estimate, per MW.

> He doesn't even mention the biggest unsolved engineering problem, removing the "ash" (Helium) that accumulates in the plasma.

And why should he? I mean, why does he have to kill the corpse a third time (after the power density and reliability arguments)?

> And that's why I believe ARC will prove Lidsky wrong.

As I said earlier, the power density figures from ARC show he was right.